Influence of pretreating methods on the score result of principal component analysis of tea

The mid-infrared spectra of six typical kinds of tea are recorded with a Fourier transform infrared spectroscopy. With different pre-treating methods, including vector normalization, derivation and smoothing, the spectra are investigated in this work. By comparing with the principal component analysis score results from different pre-treating methods, it draws conclusions: (1) The result from no pre-treating method is poor, because most scores of samples are crossing; (2) The result from vector normalization is better than that with no pre-treating method, because different kinds of tea can be distinguished basically; (3) The result from first-order derivation and smoothing is much better than the second one, because the aggregation of same samples is improved; (4) The result from vector normalization plus first-order derivative and smoothing is the best, because it makes the best aggregation of same samples and the biggest degree of decentralization between different kinds of samples, and also completely eliminates the singular points. Therefore, based on principal component analysis of infrared spectra of tea, the last pre-treating method ensures that the six kinds of tea can be distinguished accurately and effectively.